* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project
Download [pdf]
Survey
Document related concepts
Nonimaging optics wikipedia , lookup
Silicon photonics wikipedia , lookup
Surface plasmon resonance microscopy wikipedia , lookup
Optical rogue waves wikipedia , lookup
Optical tweezers wikipedia , lookup
Retroreflector wikipedia , lookup
Ellipsometry wikipedia , lookup
Harold Hopkins (physicist) wikipedia , lookup
Magnetic circular dichroism wikipedia , lookup
Interferometry wikipedia , lookup
Optical coherence tomography wikipedia , lookup
Photon scanning microscopy wikipedia , lookup
Nonlinear optics wikipedia , lookup
Transcript
Ripoll et al. Vol. 18, No. 4 / April 2001 / J. Opt. Soc. Am. A 821 Recovery of optical parameters in multiple-layered diffusive media: theory and experiments Jorge Ripoll Instituto de Ciencia de Materiales de Madrid, Consejo Superior de Investigaciones Cientı́ficas, Campus de Cantoblanco, 28049 Madrid, Spain Vasilis Ntziachristos Department of Biochemistry/Biophysics, University of Pennsylvania, Philadelphia, Pennsylvania 19104-6089 Joe P. Culver, Deva N. Pattanayak, and Arjun G. Yodh Department of Physics and Astronomy, University of Pennsylvania, Philadelphia, Pennsylvania 19104-6089 Manuel Nieto-Vesperinas Instituto de Ciencia de Materiales de Madrid, Consejo Superior de Investigaciones Cientı́ficas, Campus de Cantoblanco, 28049 Madrid, Spain Received March 30, 2000; revised manuscript received October 4, 2000; accepted October 18, 2000. Diffuse photon density waves have lately been used both to characterize diffusive media and to locate and characterize hidden objects, such as tumors, in soft tissue. In practice, most biological media of medical interest consist of various layers with different optical properties, such as the fat layer in the breast or the different layers present in the skin. Also, most experimental setups consist of a multilayered system, where the medium to be characterized (i.e., the patient’s organ) is usually bounded by optically diffusive plates. Incorrect modeling of interfaces may induce errors comparable to the weak signals obtained from tumors embedded deep in highly heterogeneous tissue and lead to significant reconstruction artifacts. To provide a means to analyze the data acquired in these configurations, the basic expressions for the reflection and transmission coefficients for diffusive–diffusive and diffusive–nondiffusive interfaces are presented. A comparison is made between a diffusive slab and an ordinary dielectric slab, thus establishing the limiting distance between the two interfaces of the slab for multiple reflections between them to be considered important. A rigorous formulation for multiple-layered (M-layered) diffusive media is put forward, and a method for solving any M-layered medium is shown. The theory presented is used to characterize a two-layered medium from transmission measurements, showing that the coefficients of scattering, s⬘ , and absorption, a , are retrieved with great accuracy. Finally, we demonstrate the simultaneous retrieval of both s⬘ and a . © 2001 Optical Society of America OCIS codes: 170.5270, 290.1990, 290.7050. 1. INTRODUCTION Diffuse photon-density waves (DPDW’s) are scalardamped traveling disturbances of light energy propagating in turbid media from an oscillating source. The present interest on them stems from their potential to locate objects hidden within these highly scattering media. The diffusion approximation1 has been shown to be reasonably good in many practical situations for describing light transport.2,3 Light propagating within the diffusion approximation has been investigated for determining the scattering and absorption coefficients of biological media as well as for formulating medical diagnoses.3–9 So far, several experiments have been performed in this area, with encouraging results,10 and basic concepts of wave propagation such as Snell’s law or diffraction from an edge have been experimentally verified.11,12 One important area in this field is the study of multiple-layered dif0740-3232/2001/040821-10$15.00 fusive media,13–23 which has been focused mainly on reflection measurements from the skin. However, reflection measurements are not sensitive to small objects embedded deep into tissue (⬃3 cm) in which case transmission measurements must also be performed. Recently, the expressions for the reflection and transmission coefficients have been presented in the case of diffusive– diffusive interfaces.24 However, other basic properties that govern these waves have not yet been established. These pertain to the reflection and refraction coefficients of their plane-wave components at the interface between a diffusive medium and an outer nondiffusive medium, where detection is performed, and to the use of these coefficients in solving multiple-layered media. This is of importance for straightforwardly determining the optical parameters. Also, these properties give us the exact value of the Green’s function in any of the layers, thus en© 2001 Optical Society of America 822 J. Opt. Soc. Am. A / Vol. 18, No. 4 / April 2001 Ripoll et al. abling us to improve the accuracy of hidden-object reconstruction within such systems. In many experimental configurations, the medium to be characterized is usually bounded between two diffusive plates. Hence, it is quite common to have a diffusive–diffusive interface between the plate and the medium and a diffusive–nondiffusive interface at the output. In most cases, to detect and characterize a hidden object within such media, this system is approximated as a one-layered medium or as a semi-infinite homogeneous medium. With the aid of the reflection and transmission coefficients for diffusive– diffusive interfaces, however, multiple-layered (Mlayered) configurations in contact with nondiffusive media can be analytically solved in terms of these coefficients, thus speeding up the computation time. In Section 2 we establish the expressions for the reflection and transmission coefficients of DPDW plane-wave components between a diffusive and a nondiffusive medium, discussing in Section 3 how an M-layered medium can be exactly solved. We consider the case of a single slab and how it compares with a dielectric slab, a consequence being that devising resonator interferometers25 is not possible for DPDW’s. In Section 4 we present the experimental setup used to verify our theory, and we present the procedure employed to fit the data in Section 5. The experimental results and simulations obtained from two different expressions for the transmitted wave, namely, a two-layered and a one-layered medium, are shown in Section 6, where we demonstrate that the twolayer expression yields accurate values of both a and ⬘s , whereas the one-layer, or semi-infinite, expressions fail to do so. In this manner, we establish a straightforward, novel, and systematic procedure for characterizing diffusive media by means of these reflection and transmission coefficients on measurements performed from a nondiffusive exterior medium. Finally, in Section 7 we present the summary and conclusions. 2. REFLECTION AND TRANSMISSION COEFFICIENTS An infinite homogeneous diffusive medium is characterized by its absorption coefficient a , the refractive index n, and the diffusion coefficient D ⫽ 1/关3 ⬘s 兴 , where s⬘ is the reduced scattering coefficient. For an intensitymodulated source at a frequency , the average intensity U(r, t) ⫽ U(r)exp(⫺it) represents the DPDW and obeys the Helmholtz equation with a wave number 0 ⫽ 关 ⫺ a /D ⫹ i /(vD) 兴 1/2, where v ⫽ c/n is the speed of light in the medium. Let a source or object be at a plane z ⫽ z s . At any plane z ⫽ constant of a homogeneous diffusive half-space, we can express the average intensity U(r) by its angular-spectrum representation of plane waves, that is, by a superposition of waves of amplitude A(K) and wave vector k ⫽ (K, q), 兩 k兩 ⫽ 0 [Refs. 26–29]: 共 UR, z 兲 ⫽ 冕 ⫹⬁ ⫺⬁ A共 K兲 exp关 iK • R ⫹ iq 共 K兲 兩 z ⫺ z s 兩 兴 dK, (1) where 兩 K兩 2 ⫹ q 2 ⫽ 02 ; i.e., K ⫽ (K x , K y ) is a real vector and q(K) ⫽ ( 02 ⫺ 兩 K兩 2 ) 1/2. For DPDW’s, since 0 is always a complex number, q(K) ⫽ q Re ⫹ iq Im is always complex, namely, q Im ⫽ 0. In Eq. (1), we choose q Re ⬎ 0 and q Im ⬎ 0 so that the field satisfies the radiation condition at infinity. From Eq. (1), we also obtain the relationships U 共 K, z 兲 ⫽ A共 K兲 exp关 iq 共 K兲 兩 z ⫺ z s 兩 兴 , J̃ n 共 K, z 兲 ⫽ ⫺D Ũ 共 K, z 兲 z (2) ⫽ ⫺iDq 共 K兲 Ũ 共 K, z 兲 , (3) where Ũ(K, z) is the Fourier transform of U(R, z). In Eq. (3), J̃ n (K, z) is the Fourier transform of the total flux density at plane z, J n (R, z), and is Fick’s law in the angular-spectrum representation. A. Diffusive–Nondiffusive Interfaces We shall now derive the reflection and transmission coefficients between diffusive and nondiffusive media. The expressions for diffusive–diffusive interfaces can be found in Ref. 24. Let us consider an isotropic and homogeneous semi-infinite diffusive medium (z ⬎ 0), separated by a plane interface at z ⫽ 0 from a nondiffusive semi-infinite half-space (z ⬍ 0). The upper diffusive medium is characterized by D 0 , a0 , n 0 , and 0 , and the lower medium is characterized by n 1 . Let the light source be located at z s ⬎ 0, with the total wave represented by U 0 ⫽ U ( i ) ⫹ U ( r ) , and U 1 ⫽ U ( t ) in the upper and lower media, respectively. The expression for U ( t ) is valid only at the interface, since propagation into the nondiffusive medium cannot be described by the diffusion approximation (see Ref. 30 for a detailed description of this problem). To find the relationship between U ( i ) , U ( r ) , and U ( t ) across the interface, and hence the expressions for the reflection and transmission coefficients Rnd and Tnd , we must apply the saltus condition at z ⫽ 0. In the diffusion approximation context, this condition is expressed by the zero flux requirement, which states that the total flux J n traversing the interface is an outward flux J ⫹ , i.e., the inward flux J ⫺ is zero,31,32 and therefore J(r) • n̂ ⫽ J n (r) ⫽ J ⫹(r). This is to be expected if there are no other light sources in the outer nondiffusive medium that could generate an inward flux. The zero-flux condition is expressed in terms of the average intensity U 0 as U 0 共 R, z 兲 兩 z⫽0 ⫽ ␣ J n 共 R, z 兲 兩 z⫽0 ⫽ ⫺D 0 ␣ 冏 U 0 共 R, z 兲 z , z⫽0 (4) where ␣ is a coefficient that takes into account refractiveindex mismatch, with the quantity D 0 ␣ ⫽ l tr␣ /3 usually referred to as the extrapolated distance,1,31,32 where l tr is the transport mean free path. In Eq. (4) we have taken the surface unit normal n̂ ⫽ ûz to be pointing outward, i.e., into the nondiffusive medium [when the surface normal is considered pointing inward, ␣ in Eq. (4) must be replaced by ⫺␣]. Substituting Eq. (3) into Eq. (4), we obtain Ũ 共 i 兲 共 K, z ⫽ 0 兲 ⫹ Ũ 共 r 兲 共 K, z ⫽ 0 兲 ⫽ ␣ J̃ n 共 K, z ⫽ 0 兲 , (5) J̃ n 共 K, z ⫽ 0 兲 ⫽ ⫺D 0 关 iq 共 K兲 Ũ 共 i 兲 共 K, z ⫽ 0 兲 ⫺ iq 共 K兲 Ũ 共 r 兲 共 K, z ⫽ 0 兲 ], (6) Ripoll et al. Vol. 18, No. 4 / April 2001 / J. Opt. Soc. Am. A where Eq. (6) represents Fick’s law and takes into account the opposite-propagation directions of U ( i ) and U ( r ) . After some simple algebra, by means of Eq. (2), Eqs. (5) and (6) reduce to 共r兲 共i兲 A 共 K兲 ⫽ Rnd共 K兲 Ũ 共 K, z ⫽ 0 兲 , 1 J n 共 K, z ⫽ 0 兲 ⫽ ␣ (8) where Rnd共 K兲 ⫽ Tnd共 K兲 ⫽ i ␣ D 0 q 共 K兲 ⫹ 1 i ␣ D 0 q 共 K兲 ⫺ 1 2i ␣ D 0 q 共 K兲 i ␣ D 0 q 共 K兲 ⫺ 1 , 共i兲 , (10) (11) Black Interface. With a particular case of diffusive– nondiffusive interface, we consider that in which the lower medium is perfectly absorbing, i.e., a black slab. The boundary condition at such an interface is U 0 共 R, z 兲 兩 z⫽0 ⫽ U 共 R, z 兲 兩 z⫽0 ⫹ U 共r兲 共 R, z 兲 兩 z⫽0 ⫽ 0, (12) which results in the following reflection and transmission coefficients for diffusive–black interfaces: Rblack共 K兲 ⫽ ⫺1, (13) Tblack共 K兲 ⫽ 0. (14) This relationship is what is expected from Eq. (12), since absorption is simply scattering with a dephase, i.e., U ( r ) (R, z ⫽ 0) ⫽ U ( i ) (R, z ⫽ 0)exp关i兴, and in the case of a perfect absorber, with the property 兩 U ( r ) 兩 ⫽ 兩 U ( i ) 兩 . Eqs. (13) and (14) are equivalent to introducing ␣ ⫽ 0 into Tnd and Rnd [see Eq. (4)]. B. Incident Field To solve any multiple-layered configuration, we need an expression for the incident field, i.e., for its angularspectrum representation. This can be obtained for a source located at z s by means of Eq. (1), namely, A共 i 兲 共 K兲 ⫽ 1 4 2 冕 ⫹⬁ ⫺⬁ U 共 i 兲 共 R, z ⫽ z s 兲 exp关 ⫺iK • R兴 dR. (15) i S0 4 D 0 2 q 共 K兲 , (16) where S 0 is the source strength. Substituting Eq. (16) into Eq. (1) yields the Weyl representation of a diverging spherical wave,27,28 which is a very useful tool for localizing and characterizing hidden objects in diffusive media.34,35 By means of Eq. (2), the expression for a point source at any plane z is Ũ 共 i 兲 共 K, z 兲 ⫽ Ũ 共 K, z ⫽ 0 兲 ⫽ Tnd共 K兲 Ũ 共 K, z ⫽ 0 兲 . 共i兲 A共 i 兲 共 K兲 ⫽ (9) Rnd and Tnd being the frequency-dependent reflection and transmission coefficients, respectively, for diffusive– nondiffusive interfaces. Both Rnd and Tnd are complex, and the sum of their moduli is not unity, but they hold the relationship Tnd(K) ⫽ Rnd(K) ⫹ 1. As can be seen in Eq. (8), we have expressed the transmission coefficient in terms of the total density flux that traverses the interface. The reason for this is that when measurements are performed from an outer nondiffusive medium, usually the measured quantity is J n , through Lambert’s cosine law25,30 and not U ( t ) . In any case, the value of U ( t ) can be directly obtained from Eq. (4), and therefore 共t兲 In the case of a point source U ( i ) (R, z ⫽ z s ) ⫽ S 0 exp关i0R兴/(4D0R), Eq. (15) reduces to (see Refs. 27, 33 for a detailed derivation) (7) Tnd共 K兲 Ũ 共 i 兲 共 K, z ⫽ 0 兲 , 823 i S0 4 D 0 2 q i 共 K兲 exp关 iq i 共 K兲 兩 z ⫺ z s 兩 兴 . (17) The incident field measured at a plane z ⭐ 0, due to a point source located at distance z s from a black boundary, is [see Eqs. (13) and (17)] Ũ 共 i 兲 共 K, z 兲 ⫽ S0 i 4 D 0 2 q i 共 K兲 兵 exp关 iq i 共 K兲 兩 z ⫺ z s 兩 兴 ⫺ exp关 iq i 共 K兲 兩 z ⫹ z s 兩 兴 其 . (18) For z ⬎ z s , Eq. (18) can be rewritten as Ũ 共 i 兲 共 K, z 兲 ⫽ S0 exp关 iq i 共 K兲 z 兴 4 D0 q i 共 K兲 2 sinh关 iq i 共 K兲 z s 兴 . (19) We should state that the correct way to represent the source is to include an angular-dependent, or dipolar, term as in Ref. 36. Nevertheless, since this effect decays at long distances from the source, we will from now on use Eqs. (17) and (18) to model the incident field. 3. MULTIPLE-LAYERED MEDIA We next solve any multiple-layered (M-layered) system without any approximation. For the sake of clarity, we shall first address the simple system consisting of a slab that is equivalent to a three-layered medium. This is done for two reasons: First, the expression for a slab will be very useful, since in many cases, a system of multiple layers can be approximated by a system of slabs with no multiple reflections between them; second, a great deal of information, such as the limiting depth at which information from an object can be recovered or the multiplereflection contribution, can be extracted. A. Expression for a Slab Let us address the configuration depicted in Fig. 1, namely, a slab of width L, located at 0 ⬍ z ⬍ L. At z ⬎ L there is a semi-infinite homogeneous medium of parameters D 0 , a0 , n 0 , and 0 , with a source located at z s . At z ⬍ 0 we have a semi-infinite homogeneous medium of parameters D 2 , a2 , n 2 , and 2 . The slab has D 1 , a1 , n 1 , and 1 as parameters. We shall consider the normal n̂ at the interfaces to be pointing in the ⫺z direction, i.e., n̂ ⫽ (0,0,⫺1). There are two ways of solving this configuration. The first method consists of introducing the boundary conditions at each of the interfaces and solving the linear system of equations by following a 824 J. Opt. Soc. Am. A / Vol. 18, No. 4 / April 2001 Ripoll et al. Equations (23) and (24) correspond to a dielectric slab.25 For a gain medium, they represent the equations for a Fabry–Perot laser cavity,37 and R10R12 exp关2iq1L兴 ⫽ 1 represents the oscillation condition. In Eqs. (23) and (24), the denominator 1 ⫺ R10R12 exp关2iq1L兴 takes into account the multiple reflections between the interfaces of the slab. Since in general for DPDW’s R10R12 exp关2iq1L兴 Ⰶ 1, the geometric progression can be truncated to first order, i.e., Rslab ⯝ R01 ⫹ T01 exp关 2iq 1 L 兴 R12T10 , Fig. 1. Configuration for a slab where three different zones are distinguished, with different diffusive parameters: z ⭓ L, 0 ⭐ z ⭐ L, and z ⭐ 0. scheme similar to that presented in Refs. 15, 17, 19, and 23 for a two-layered medium. This is the method that will be employed for solving M-layered systems. The second method consists of successively adding multiple reflections and transmissions at the interfaces, and since it is physically illustrative we shall use it to derive the expressions for a slab. The total wave in the three regions of Fig. 1 is Ũ ( i ) ⫹ A exp关iq0(z ⫺ L)兴, z ⬎ L;B exp关iq1(L ⫺ z)兴 ⫹ C exp关iq1z兴, L ⬎ z ⬎ 0; and D exp关⫺iq2z兴, z ⬍ 0. Considering all multiple reflections from the interfaces, the total reflection and transmission coefficients for the slab are25,37 Rslab ⫽ R01 ⫹ T01R12 exp关 2iq 1 L 兴 T10 ⫹ T01R12 ⫻ exp关 2iq 1 L 兴 R10R12 ⫻ exp关 2iq 1 L 兴 T10 ⫹ ..., T slab T slab ⯝ T01 exp关 iq 1 L 兴 T12 . These approximations are very accurate, even for low values of L, as shown in Fig. 2. In this figure one sees that for values of L greater than 3 cm the reflection from the slab is simply R01 . Therefore R10R12 exp关2iq1L兴 represents a way of determining the distance at which multiple reflections between the walls are important. As a rule of thumb, we can use the following criterion: Assuming 0.1% noise in the measurements, the limiting value of exp关2iq1L兴 for absence of multiple reflection is 0.1% at best. This means L ⫽ ⫺log关10⫺3 兴 /2 0 for the case ⫽ 0, K ⫽ 0. Therefore the limiting value of L is approximately L limit ⬃ 3 共 D 1 / a1 兲 1/2. (25) ⫺1 ⬘ ⫽ 10 cm we show that For example, in Fig. 2, for s1 ⬘ ⫽ 20 cm⫺1, we obtain L limit L limit ⬃ 3 cm, and for s1 ⬃ 2.5 cm. For a dielectric slab, the condition R10R12 exp关2iq1L兴 ⬃ 1 conveys the existence of transmission and reflection (20) ⫽ T01 exp关 iq 1 L 兴 T10 ⫹ T01R12 exp关 2iq 1 L 兴 R10 ⫻ exp关 iq 1 l 兴 T10 ⫹ ..., (21) where Rij and Tij are the reflection and transmission coefficients going from medium i to medium j (Ref. 24). Equations (20) and (21) contain a geometric progression, and therefore A, B, C, and D are extracted as 再 A⫽ R01 ⫹ T01 exp关 2iq 1 L 兴 R12T10 1 ⫺ R10R12 exp关 2iq 1 L 兴 T01 B⫽ 1 ⫺ R10R12 exp关 2iq 1 L 兴 T01 exp关 iq 1 L 兴 R12 C⫽ 1 ⫺ R10R12 exp关 2iq 1 L 兴 T01 exp关 iq 1 L 兴 T12 D⫽ 1 ⫺ R10R12 exp关 2iq 1 L 兴 冎 Ũ 共 i 兲 , Ũ 共 i 兲 , Ũ 共 i 兲 , Ũ 共 i 兲 . (22) From Eqs. (22) the reflection and transmission coefficients for a slab are defined as R slab共 K兲 ⫽ R01 ⫹ T 共 K兲 ⫽ slab T01 exp关 2iq 1 L 兴 R12T10 1 ⫺ R10R12 exp关 2iq 1 L 兴 T01 exp关 iq 1 L 兴 T12 1 ⫺ R10R12 exp关 2iq 1 L 兴 . , (23) (24) Fig. 2. Amplitude of Rslab and T slab at K ⫽ 0 for different val⬘ ⫽ 10 cm⫺1, ⫽ 0 (solid ues of the slab width L for the cases s1 ⫺1 ⬘ ⫽ 20 cm , ⫽ 0 (solid circles); s1 ⬘ ⫽ 20 cm⫺1, curves); s1 ⬘ ⫽ s2 ⬘ ⫽ 5 cm⫺1, ⫽ 200 MHz (open circles). In all cases s0 a0 ⫽ a1 ⫽ a2 ⫽ 0.025 cm⫺1, n 0 ⫽ n 1 ⫽ n 2 ⫽ 1.333. Ripoll et al. Vol. 18, No. 4 / April 2001 / J. Opt. Soc. Am. A 825 conclude that Fabry–Perot resonators are not possible for DPDW’s. Additional useful information contained in Eq. (23) pertains to the maximum distance between the slab walls from which parameters from medium 2 can be extracted when reflection measurements from medium 0 are performed, since all the information from medium 2 is contained in R12 exp关2iq1L兴. In this case we can also use Eq. (25), thus concluding that in reflection measurements, it is not possible to characterize an object buried at a distance larger than L limit from the first interface. By contrast, as seen from Eq. (24), this does not occur in transmission measurements, since information of all the media is present in the wave field. B. Solving Multiple-Layered Media Let us address an M-layered system, as depicted in Fig. 4, where n in corresponds to the refractive index of the nonscattering medium that bounds the medium containing the source and n out stands for the refractive index of the nonscattering medium where detection is performed. An equivalent solution for multiple-layered dielectric media can be found in Ref. 25. In any of the jth inner media, the total field for each frequency component is U j ⫽ Aj exp关 iq j 共 z j ⫺ z 兲兴 ⫹ Bj exp关 iq j 共 z ⫺ z j⫹1 兲兴 , z j ⭐ z ⭐ z j⫹1 , slab 关 M̂ 兴 M ⫻ M • 关 X̂ 兴 M ⫻ 1 ⫽ 关 Ŷ 兴 M ⫻ 1 , resonant peaks. However, this requires values of R10 and R12 close to unity. This is not the case for a diffusive 关 M̂ 兴 ⬅ 冤 (27) where f in ⫺g inE 1 0 0 0 0 a 12E 1 b 12 ⫺1 ⫺E 2 0 0 q 1D 1E 1 ⫺q 1 D 1 ⫺q 2 D 2 q 2D 2E 2 0 0 0 0 a 23E 2 b 23 ⫺1 ⫺E 3 0 0 q 2D 2E 2 ⫺q 2 D 2 ⫺q 3 D 3 q 3D 3E 3 0 0 0 0 ] ] ] ] ] ] ] ] 0 0 0 0 0 0 slab. Also, we must take into consideration that q 1 is always complex, and therefore exp关iq1L兴 represents a lossy medium as it decays with L. The behavior of Rslab and T slab versus the modulation frequency f ⫽ /2 is shown in Fig. 3, where we see that for f ⬍ 1 MHz very little difference is observed in the dc ( ⫽ 0) case. For high values of , we must recall that the imaginary part of q 1 , and therefore the decay of exp关2iq1L兴, grows as 1/2. Nevertheless, in Fig. 3 we distinguish maxima near f ⫽ 200 MHz. These maxima correspond to maximum values of exp关2iq1L兴 and therefore cannot be considered as resonant peaks as for the dielectric slab. We therefore (26) where z j is the position of jth interface, i.e., z j k⫽j ⫽ 兺 k⫽1 L k , L k being the width of medium k. On introducing the saltus conditions for each of the interfaces, taking into consideration that the first and last interfaces are diffusive–nondiffusive, we get to the following set of equations: Fig. 3. Amplitude and phase of R and T at K ⫽ 0 for dif⬘ ferent values of the modulation frequency for the cases s1 ⬘ ⫽ 10 cm⫺1 [Rslab (solid curves) and T slab (dotted curves)], s1 ⫽ 20 cm⫺1 [Rslab (solid circles) and T slab (open circles)]. In all ⬘ ⫽ s2 ⬘ ⫽ 5 cm⫺1, a0 ⫽ a1 ⫽ a2 ⫽ 0.025 cm⫺1, n 0 cases s0 ⫽ n 1 ⫽ n 2 ⫽ 1.333. slab ¯ ¯ ¯ ¯ 0 0 0 0 0 0 0 0 ¯ 0 0 ¯ ] ] ⫺q M D M q MD ME M ⫺g outE M f out ¯ 冥 , (28) Fig. 4. Multiple-layered configuration of M slabs, where n in and n out are the refractive indices of the input and output media, respectively, with both being nondiffusive. In all cases, the normal to each interface is considered to point from n in into n out , i.e., along the propagation direction of the incident wave. 826 J. Opt. Soc. Am. A / Vol. 18, No. 4 / April 2001 关 X̂ 兴 ⬅ 冉冊 冉 A1 B1 ] ; AM BM 关 Ŷ 兴 ⬅ 冊 Ripoll et al. g inŨ 共 i 兲 共 K, z ⫽ 0 兲 ⫺a 12Ũ 共 i 兲 共 K, z ⫽ 0 兲 , 0 ] 0 (29) and E j ⫽ exp关 iq j L j 兴 , (30) a jk ⫽ 共 n k /n j 兲 ⫹ iCjk D j q j , 2 b jk ⫽ 共 n k /n j 兲 2 ⫺ iCjk D j q j . Fig. 5. Equations (31) are the coefficients for the boundary conditions between the jth and k ⫽ j ⫹ 1 diffusive media, which in the case n j ⫽ n k have the values a jk ⫽ b jk ⫽ 1 (see Refs. 24, 38). For the input and output interfaces we have defined f in ⫽ i ␣ inq 1 D 1 ⫺ 1; f out ⫽ i ␣ outq M D M ⫺ 1; g in ⫽ i ␣ inq 1 D 1 ⫹ 1, Experimental setup. (31) (32) g out ⫽ i ␣ outq M D M ⫹ 1, (33) which when the input interface is black ( ␣ in ⫽ 0), yield f in ⫽ ⫺1, g in ⫽ 1. Notice that g x /f x ⫽ Rnd [see Eq. (9)]. As seen from Eq. (27), after solving the linear system of equations we are left with an M ⫻ 1 matrix that represents the values of Aj , Bj for each frequency component. That is, Eq. (27) must be solved for each frequency component. Once these Aj , Bj coefficients are found, the total field can be built inside any of the slabs by Eq. (26). 4. EXPERIMENTAL SETUP To test the theory, we measured a two-layer system composed of an intralipid layer and a resin layer, as depicted in Fig. 5. This setup is one that can be used to perform breast measurements in the cw regime, i.e., no modulation frequency, in which case the breast is introduced in intralipid. The resin was a polyester resin with a TiO2 suspension39 of length L 1 ⫽ 2.17 cm with optical param⬘ ⫽ 10 cm⫺1, a1 ⫽ 0.017 cm⫺1, and n 1 ⫽ 1.5. eters s1 The intralipid used was an Intralipid (Kabi Pharmacia, Clayton, North Carolina) emulsion, which is a polydisperse suspension of fat particles ranging in diameter from 0.1 to 1.1 m that served as the scattering background medium. This intralipid layer had a length L 0 ⬘ , a0 , and n 0 , for which ⫽ 4 cm and parameters s0 ⬘ , were used, namely, three different values of s0 ( i⫽1, 2, 3 ) s0 ⫽ 5 cm⫺1, 10 cm⫺1, and 20 cm⫺1. For each of (i) these sets s0 , five values of a0 where introduced, ( j⫽1,...,5 ) namely, a0 ⫽ 0.025 cm⫺1, 0.050 cm⫺1, 0.075 cm⫺1, ⫺1 0.1 cm , and 0.125 cm⫺1, constituting a total of 15 differ(i) ( j) , a0 pairs. In all cases the refractive index of ent s0 the intralipid was n 0 ⫽ 1.333. To carry out the experiment, the necessary amount of intralipid was mixed with ⬘ desired. Then the value water to obtain the value of s0 of a0 was changed by introducing black India ink (3080-4 KOH-I-NOOR Inc., Bloomsbury, New York) into (i) the mixture, performing a measurement for each s0 , ( j) a0 pair. As shown in Fig. 5, the system had side walls of black PVC. Measurements were performed at the exit of the resin layer. This means that the configuration had three interfaces, namely, black–intralipid, intralipid–resin, and resin–air interfaces. The basin had a height h ⫽ 15 cm, width w ⫽ 25 cm, and total length L ⫽ 7 cm (including the black PVC walls). In spite of the finite size of the basin, we consider the system to be infinite in the XY plane, and thus the expressions put forward in Section 2 will be employed. Illumination was accomplished by using an optical fiber of NA ⫽ 0.36, which emitted light of constant intensity at ⫽ 786 nm with a laser diode power ⬃3 mW, located at the rear of the basin at (x s ⫽ 8.6 cm, y s ⫽ 8.2 cm, z s ⫽ 0). We shall model this source by a point source located at rs ⫽ (x s , y s , z s ⫽ l tr), in the cw regime, i.e., ⫽ 0 (constant illumination). Measurements were taken with a liquid-nitrogencooled, 16-bit CCD array (Princeton Instruments), which had a resolution of 330 ⫻ 1100 pixels that were 24 m in linear dimension, each pixel representing an area dx ⫻ dy ⫽ 0.0287 cm ⫻ 0.0287 cm. The CCD was focused by means of lenses on the exit surface of the resin, i.e., at z ⫽ L 0 ⫹ L 1 ⫽ 6.17 cm. 5. DATA ANALYSIS The two-layered system shown in Fig. 5 is numerically solved by using Eqs. (27)–(29), with ␣ in ⫽ 0, and ␣ out ⯝ 7.25, corresponding to n out ⫽ 1, n 1 ⫽ 1.5. The three interfaces of the experimental setup (see Fig. 5), two of which are diffusive–nondiffusive, yield a linear system of four equations and four unknowns. An approximation to the system of Eqs. (27)–(29) can be introduced by using Eq. (25), since we expect no contribution from multiple reflections between the walls of the intralipid, where L 0 ⫽ 4 cm. Also, the highest value of L limit is L limit ⬃ 5 cm, for the ( s0 ⫽ 5 cm⫺1, a0 ⫽ 0.025 cm⫺1) case, taking into consideration that one of the interfaces is black. Hence we simply assume an incident wave coming from the intralipid whose wave function is represented by Eq. (19), which interacts with a slab (i.e., the resin block), as described by Eq. (24). Therefore, with Eq. (11), the expected average intensity at the exit of the resin, z d ⫽ L 0 ⫹ L 1 ⫽ 6.17 cm, is Ũ 2layer共 K, z d 兲 ⫽ S0 exp关 iq 0 共 K兲 L 0 兴 4 D0 q 0 共 K兲 2 ⫻ sinh关 iq 0 共 K兲 z s 兴 T slab resin共 K 兲 , (34) Ripoll et al. Vol. 18, No. 4 / April 2001 / J. Opt. Soc. Am. A slab where T resin is given by Eq. (24) if medium 2 is replaced by a nonscattering medium, i.e., by replacing R12 with Rnd , and T12 with Tnd [see Eqs. (9) and (10)]. We remark that Eq. (34) has been contrasted with the exact solution for Eq. (27), and no noticeable differences were found. To simulate the case in which we use the setup of Fig. 5 for breast characterization, we assume that the parameters of one of the media—namely, the resin layer—are known, and we must therefore fit the values of the intralipid. For the fitting functions, we shall use two cases, namely, a two-layered medium [Eq. (34)] and a onelayered medium. The expression for the one-layered medium is obtained by considering that parameters of the resin (see Fig. 5) are the same as the intralipid, and therefore Ũ 1layer共 K, z d 兲 ⫽ S0 exp关 iq 0 共 K兲共 L 0 ⫹ L 1 兲兴 4 2D 0 q 0 共 K兲 ⫻ sinh关 iq 0 共 K兲 z s 兴 Tnd共 K兲 , 827 (i) . The fitting procedure employed is the value of s0 steepest-descent method, in which we shall always use a0 ⫽ 0.1 cm⫺1, s0 ⫽ 3.33 cm⫺1 as initial values (see Ref. 17 for comparison between different fitting procedures). The function to minimize is f 关 共s0i 兲 , 共a0j⫽1,...,5 兲 兴 ⫽ 兺 储 Ū 共 i, j⫽1,...,5 兲 共 K兲 theory K i, j⫽1,...,5 兲 ⫺ Ū 共data 共 K兲储 2 , (36) where the subindex theory stands for either 2 layer or 1 layer [see Eqs. (34) and (35)], depending on the expression i, j⫽1,...,5 ) used to fit, and Ū (( ... represent the expressions for ) (i) ( j⫽1,...,5 ) the set 关 s0 , a0 兴 normalized to the K ⫽ 0 value corresponding to ( s0 ⫽ 5 cm⫺1, a0 ⫽ 0.025 cm⫺1). To (35) where Tnd is the transmission coefficient for the intralipid–air interface. For both the two-layer and one(i) layer expressions we shall use z s ⫽ l tr ⫽ 3D 0 ⫽ 1/ s0 . ( j⫽1,...,5 ) ( i⫽1, 2, 3 ) To fit for a0 and s0 , we first perform the ( i, j ) two-dimensional Fourier transform Ũ data (K) of the CCD ( i, j ) (i) ( j) measurements U data (R) corresponding to the ( s0 , a0 ) ⫺1 pair. We shall use a cutoff frequency Kcut ⫽ 1.2 cm to separate noise from the data. Then we normalize all ( i⫽1, 2, 3 ) ( j⫽1,...,5 ) , a0 关 s0 兴 sets to the value corresponding to ( i⫽1 ) ( j⫽1 ) [ s0 ⫽ 5 cm⫺1, a0 ⫽ 0.025 cm⫺1] at K ⫽ 0, finally ( i⫽1, 2, 3 ) ( j⫽1,...,5 ) fitting for both s0 and a0 . In this way, we make the assumption that we only have one reference measurement for which the values s0 and a0 are known. In all cases, the only restriction that we impose ( j⫽1,...,5 ) is that each set a0 is associated with the same Fig. 6. Fitted values of Ū 2slab(K) (solid circles) and Ū 1slab(K) (open triangles), compared with the data values Ū data(K) (solid ⬘ ⫽ 10 cm⫺1, a0 curve) versus the frequency K, for the case s0 ⫽ 0.075 cm⫺1. See Eqs. (34) and (35). ⬘ ⫽ 5 cm⫺1, (b) s0 ⬘ ⫽ 10 cm⫺1, (c) s0 ⬘ ⫽ 20 cm⫺1, and the Fig. 7. Fitted values with use of the two-layer expression [Eq. (34)]: (a) s0 ⫺1 ⫺1 ⫺1 ⬘ ⫽ 5 cm , (e) s0 ⬘ ⫽ 10 cm , ( f ) s0 ⬘ ⫽ 20 cm . Symbols and solid curves, the expected and one-layer expression [Eq. (35)]: (d) s0 the retrieved values, respectively. 828 J. Opt. Soc. Am. A / Vol. 18, No. 4 / April 2001 Ripoll et al. define the accuracy of each expression of retrieving the parameters, we define the average error for the retrieval of a0 in units of cm⫺1 as ⌬ a0 ⫽ 1 5 5 兺 found expected 兩 a0 共 j 兲 ⫺ a0 共 j 兲兩 关 cm⫺1 兴 , (37) Table 1. Comparison of the Fitted Values Obtained with Eqs. (34) and (35), Two-Layered Medium and One-Layered Medium, Respectivelya expected s0 (cm⫺1) 2layer s0 (cm⫺1) 2layer ⌬ a0 (cm⫺1) 1layer s0 (cm⫺1) 1layer ⌬ a0 (cm⫺1) 5.0 10.0 20.0 4.1 7.0 16.8 ⫾0.022 ⫾0.025 ⫾0.004 9.9 4.7 7.3 ⫾0.059 ⫾0.032 ⫾0.035 j⫽1 where j ⫽ 1...5) ⫽ (0.025 cm⫺1,...,0.125 cm⫺1) found are the expected values of a0 , and a0 ( j ⫽ 1...5) are those fitted from Eqs. (34) and (35). expected a0 ( a 6. RESULTS AND DISCUSSION Figure 6 shows the fitted curves of Ū 2layer and Ū 1layer versus the frequency K, contrasted with the data set Ū data for ⬘ ⫽ 10 cm⫺1, a0 ⫽ 0.075 cm⫺1). As shown, the case ( s0 both expressions Ū 2layer and Ū 1layer yield close fits to the experimental data Ū data , even at high values of the frequency K ⯝ 1.2 cm⫺1. However, as will be seen in the refound found constructions, the values of ( s0 , a0 ), which yield the curves in Fig. 6 with use of the one-layer expression, Eq. (35), are far from the expected values but are very near the expected values when using the two-layer expression, Eq. (34). The results obtained from the fitting procedure are presented in Fig. 7 and Table 1, which (i) ( j) show that the values of both s0 and a0 are retrieved with remarkable accuracy with the two-layer expression. On the other hand, the one-layer expression yields accept( j) able reconstruction values for a0 [see Figs. 7(d)–7( f )] (i) but fails to reconstruct s0 (see Table 1). We would like to draw attention to the values of ⌬ a0 shown in Table 1. In this table, we show that the error of retrieval of a0 can amount to as high as ⫾0.025 cm⫺1 for the two-layer expression. If we look at Fig. 7(b), which shows the refound trieved values of a0 versus the expected values of expected a0 , we see that even in the case in which ⌬ a0 ⫽ ⫾0.025 cm⫺1 (which corresponds to s0 ⫽ 10 cm⫺1), ( j) the relative values of a0 are correct. That is, if the exact value of one of them is known, then all the others lie correctly in place, yielding a very small value of ⌬ a0 . Also, as shown in Table 1, the retrieved values of s0 are all within a 30% error for the two-layer expression, whereas for the one-layer expression they can be as high as 70%. The same calculations were performed with the expression for the semi-infinite configuration, Eq. (19), and in all cases the values of ⌬ a0 were of the order of ⬘ were ⫾0.04 cm⫺1, and the errors of retrieval for s0 ⬃70%. We emphasize that when performing transmission measurements in the cw regime, in most cases the ⬘ and a0 cannot be retrieved simultaneously values of s0 from independent measurements unless we choose initial fitting values very close to the real ones. This is due to uncertainties in the measurements and in the optical properties, resulting in a large number of local minima that occur whenever the coefficients D 0 and a0 have the found same relative values, i.e., D 0found/ a0 expected expected ⫽ D0 / a0 . Even so, we have been able to retrieve both values from independent measurements for ⬘ ⫽ 5 cm⫺1 with errors lower than 15%, the case with s0 but we have found deviations in the order of 100% for the ⬘ ⫽ 10 cm⫺1, 20 cm⫺1 cases. Since results were not s0 2,1layer s0 are the fitted values for s0 using the two-layer and one-layer 2,1layer is the average error of retrieval of expressions, respectively, and ⌬ a0 a0 , Eq. (37). general, independent fits are not presented. As shown in Refs. 17 and 18 this does not occur when measuring spatially resolved reflection in the cw regime. Also, when performing reflection measurements in the frequency or time domains, it has been shown experimentally that one can also fit for the slab width L 0 with great accuracy, but that fitting for all the optical parameters of both resin and intralipid yields errors higher than 200% (Ref. 23). 7. CONCLUSIONS We have put forward the expressions for the reflection and transmission coefficients at diffusive–nondiffusive interfaces. These coefficients, together with those already established for diffusive–diffusive interfaces,24 can be used to find the exact solution for any multiple-layered configuration, in either the time domain or the frequency domain, and hence to retrieve the optical parameters. In addition to being useful in characterizing multiplelayered media or in finding the exact solution of complex forward problems, these coefficients also constitute a powerful tool to generate more-accurate Green’s functions that can be used to detect and characterize hidden objects within such media.8,40,41 With these expressions we have shown that any multiple-layered system can be rigorously solved at a very low computational cost. We have put forward the simple example of a slab for which the reflection and transmission coefficients have been derived. The expressions for a slab can be used to obtain the limiting distance at which multiple reflections between two interfaces exist and also to give information about using reflection measurements to characterize objects buried at a certain distance under an interface. A comparison with a dielectric slab has been presented, demonstrating that in the case of diffusive waves no interference resonant peaks can be observed and therefore that interferometers based on plane parallel plates, such as the Fabry–Perot, cannot be operated for diffusive waves. The expression for a slab is also useful, since in most cases multiplelayered media can be approximated to a series of slabs with no multiple reflections between them. Once we have established the method for solving multiple-layered media, we have characterized a twolayered system from experimental data for 15 different ( s⬘ , a ) pairs. In this characterization procedure we have demonstrated that using the expression for a twolayered system yields better reconstructed values for both ⬘s and a than the one-layer, or semi-infinite expression. Ripoll et al. We have shown that by means of the two-layer expression obtained from the reflection and transmission coefficients, we can reconstruct both the values of ⬘s and a from transmission measurements without any a priori information on their values by using one reference measurement. In all cases the values of ⬘s and a were retrieved within a ⬃30% error. Vol. 18, No. 4 / April 2001 / J. Opt. Soc. Am. A 15. 16. 17. ACKNOWLEDGMENTS We thank an anonymous reviewer for valuable comments. This research has been partially supported by the Fundación Ramón Areces. J. Ripoll acknowledges a grant from the Ministerio de Educación y Cultura. V. Ntziachristos acknowledges National Institutes of Health grant CA60182. A Fortran program for solving exactly any M-layered system is available on request at [email protected]. Corresponding author J. Ripoll can be reached at Institute of Electronic Structure and Laser, FORTH, P.O. Box 1527, GR-71110 Heraklion, Greece, or by e-mail at [email protected]. 18. 19. 20. 21. 22. REFERENCES AND NOTES 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. A. Ishimaru, Wave Propagation and Scattering in Random Media (Academic, New York, 1978), Vol. 1. M. S. Patterson, B. Chance, and B. C. Wilson, ‘‘Timeresolved reflectance and transmittance for the noninvasive measurement of tissue optical properties,’’ Appl. Opt. 28, 2331–2336 (1989). A. Yodh and B. Chance, ‘‘Spectroscopy and imaging with diffusing light,’’ Phys. Today 48, 38–40 (1995). H. Jiang, K. D. Paulsen, U. L. Osterberg, B. W. Pogue, and M. S. Patterson, ‘‘Simultaneous reconstruction of optical absorption and scattering maps in turbid media from nearinfrared frequency-domain data,’’ Opt. Lett. 20, 2128–2130 (1995). S. R. Arridge, P. Van Der Zee, M. Cope, and D. T. Delpy, ‘‘Reconstruction methods for near-infrared absorption imaging,’’ in Time-Resolved Spectroscopy and Imaging of Tissues, B. Chance and A. Katzir, eds., Proc. SPIE 1431, 204– 215 (1991). E. B. de Haller, ‘‘Time-resolved transillumination and optical tomography,’’ J. Biomed. Opt. 1, 7–17 (1996). S. Fantini, S. A. Walker, M. A. Franceschini, M. Kaschke, P. M. Schlag, and K. T. Moesta, ‘‘Assessment of the size, position, and optical properties of breast tumors in vivo by noninvasive optical methods,’’ Appl. Opt. 37, 1982–1989 (1998). V. Ntziachristos, X. Ma, and B. Chance, ‘‘Time-correlated single photon counting imager for simultaneous magnetic resonance and near-infrared mammography,’’ Rev. Sci. Instrum. 69, 4221–4233 (1998). S. R. Arridge, ‘‘Optical tomography in medical imaging,’’ Inverse Probl. 15, R41–R93 (1999). J. G. Fujimoto and M. S. Patterson, eds., Advances in Optical Imaging and Photon Migration, Vol. 21 of OSA Trends in Optics and Photonic Series (Optical Society of America, Washington, D.C., 1998). M. A. O’Leary, D. A. Boas, B. Chance, and A. G. Yodh, ‘‘Refraction of diffuse photon density waves,’’ Phys. Rev. Lett. 69, 2658–2661 (1992). J. B. Frishkin and E. Gratton, ‘‘Propagation of photondensity waves in strongly scattering media containing an absorbing semi-infinite plane bounded by a straight edge,’’ J. Opt. Soc. Am. A 10, 127–140 (1993). M. Keijzer, W. M. Star, and P. R. M. Storchi, ‘‘Optical diffusion in layered media,’’ Appl. Opt. 27, 1820–1824 (1988). J. M. Schmitt, G. X. Zhou, and E. C. Walker, ‘‘Multilayer 23. 24. 25. 26. 27. 28. 29. 30. 31. 32. 33. 34. 35. 36. 37. 38. 829 model of photon diffusion in skin,’’ J. Opt. Soc. Am. A 7, 2141–2153 (1990). I. Dayan, S. Havlin, and G. H. Weiss, ‘‘Photon migration in a two-layer turbid medium: a diffusion analysis,’’ J. Mod. Opt. 39, 1567–1582 (1992). A. H. Hielscher, H. Liu, and B. Chance, ‘‘Time-resolved photon emission from layered turbid media,’’ Appl. Opt. 35, 2221–2227 (1996). G. Alexandrakis, T. J. Farrell, and M. S. Patterson, ‘‘Accuracy of the diffusion approximation in determining the optical properties of a two-layer turbid medium,’’ Appl. Opt. 37, 7401–7409 (1998). T. J. Farrell, M. S. Patterson, and M. Essenpreis, ‘‘Influence of layered tissue architecture on estimates of tissue optical properties obtained from spatially resolved diffuse reflectometry,’’ Appl. Opt. 37, 1958–1972 (1998). A. Kienle, M. S. Patterson, N. Dögnitz, ‘‘Noninvasive determination of the optical properties of two-layered turbid media,’’ Appl. Opt. 37, 779–791 (1998). A. Kienle, T. Glanzmann, G. Wagnieres, ‘‘Investigation of two-layered turbid media with time-resolved reflectance,’’ Appl. Opt. 37, 6852–6862 (1998). L. O. Svaasand, T. Spott, J. B. Frishkin, T. Pham, B. J. Tromberg, and M. W. Berns, ‘‘Reflectance measurements of layered media with diffuse photon-density waves: a potential tool for evaluating deep burns and subcutaneous lesions,’’ Phys. Med. Biol. 44, 801–813 (1999). G. Alexandrakis, T. J. Farrell, and M. S. Patterson, ‘‘Monte Carlo diffusion hybrid model for photon migration in a twolayer turbid medium in the frequency domain,’’ Appl. Opt. 39, 2235–2244 (2000). T. H. Pham, T. Spott, L. O. Svaasand, and B. J. Tromberg, ‘‘Quantifying the properties of two-layer turbid media with frequency-domain diffuse reflectance,’’ Appl. Opt. 39, 4733– 4745 (2000). J. Ripoll and M. Nieto-Vesperinas, ‘‘Reflection and transmission coefficients for diffuse photon-density waves,’’ Opt. Lett. 24, 796–798 (1999). M. Born and E. Wolf, Principles of Optics, 6th ed. (Pergamon, New York, 1993). J. W. Goodman, Introduction to Fourier Optics (McGrawHill, New York, 1968). L. Mandel and E. Wolf, Optical Coherence and Quantum Optics (Cambridge U. Press, Cambridge UK, 1995). M. Nieto-Vesperinas, Scattering and Diffraction in Physical Optics (Wiley–Interscience, New York, 1991). J. Ripoll, M. Nieto-Vesperinas, and R. Carminati, ‘‘Spatial resolution of diffuse photon density waves,’’ J. Opt. Soc. Am. A 16, 1466–1476 (1999). J. Ripoll, S. R. Arridge, H. Dehghani, and M. NietoVesperinas, ‘‘Boundary conditions for light propagation in diffusive media with nonscattering regions,’’ J. Opt. Soc. Am. A 17, 1671–1681 (2000). R. C. Haskell, L. O. Svaasand, T. Tsay, T. Feng, M. S. McAdams, and B. J. Tromberg, ‘‘Boundary conditions for the diffusion equation in radiative transfer,’’ J. Opt. Soc. Am. A 11, 2727–2741 (1994). R. Aronson, ‘‘Boundary conditions for diffusion of light,’’ J. Opt. Soc. Am. A 12, 2532–2539 (1995). A. Banos, Dipole Radiation in the Presence of a Conducting Half-Space (Pergamon, New York, 1966). X. De Li, T. Durduran, A. G. Yodh, B. Chance, and D. N. Pattanayak, ‘‘Diffraction tomography for biochemical imaging with diffuse-photon density waves,’’ Opt. Lett. 22, 573– 575 (1997). T. Durduran, J. P. Culver, M. J. Holboke, X. D. Li, L. Zubkov, B. Chance, D. N. Pattanayak, and A. G. Yodh, ‘‘Algorithms for 3D localization and imaging using near-field diffraction tomography with diffuse light,’’ Opt. Exp. 4, 247–262 (1999). D. N. Pattanayak and A. G. Yodh, ‘‘Diffuse optical 3D-slice imaging of bounded turbid media using a new integrodifferential equation,’’ Opt. Exp. 4, 231–240 (1999). A. Yariv, Introduction to Optical Electronics, 2nd ed. (Holt, Rinehart & Winston, New York, 1976). J. Ripoll and M. Nieto-Vesperinas, ‘‘Index mismatch for dif- 830 39. J. Opt. Soc. Am. A / Vol. 18, No. 4 / April 2001 fuse photon-density waves both at flat and rough diffusediffuse interfaces,’’ J. Opt. Soc. Am. A 16, 1947–1957 (1999). Detailed information on the components and on how to build the resin can be found at http:// www.med.upenn.edu/⬃oisg/oisg.html. Ripoll et al. 40. 41. V. Ntziachristos, X. Ma, A. G. Yodh, and B. Chance, ‘‘Multichannel photon counting instrument for spatially resolved near-infrared spectroscopy,’’ Rev. Sci. Instrum. 70, 193–201 (1999). V. Ntziachristos, B. Chance, and A. G. Yodh, ‘‘Differential diffuse optical tomography,’’ Opt. Exp. 5, 230–242 (1999).